UGA Codon Position Affects the Efficiency of Selenocysteine Incorporation into Glutathione Peroxidase-1*

A UGA codon and a selenocysteine insertion sequence in the 3′-untranslated region are the only established mRNA elements necessary for selenocysteine (Sec or U) incorporation during translation. These two elements, however, do not universally confer efficient Sec incorporation. The objective of this study was to systematically examine the effect of UGA codon position on efficiency of Sec insertion. In a glutathione peroxidase-1 (F-GPX1) expression vector, the UGA at the native position (U47) was mutated to a cysteine codon, and codons for Ser-7, Ser-12, Ser-18, Ser-29, Ser-45, Ser-93, Cys-154, Val-172, Ser-178, and Ser-195 were individually mutated to UGA and transiently expressed in COS-7 cells. 75Se incorporation at the 11 positions was 31, 72, 54, 105, 90, 100, 146, 135, 13, 11, and 43%, respectively, of 75Se incorporation at U47, suggesting that Sec is more efficiently incorporated at UGA codons positioned in the middle of the coding region rather than close to the 5′ or 3′ ends. Ribonuclease protection showed that these differences were not due to differences in mRNA level. When the green fluorescence protein (GFP) coding region was placed in-frame at the 5′ or 3′ ends of the coding region in F-GPX1 to produce chimeric 50–51-kDa GFP/GPX1 proteins, Sec incorporation at UGA codons, formerly close to the 5′ or 3′ ends, was increased to levels comparable to the UGA at U47. Insertion of GFP after the UAA-stop was just as effective in increasing Sec insertion efficiency as GFP inserted before the stop. These studies used a recombinant expression model that incorporated Sec at non-native UGA codons at rates equal to those of endogenous glutathione peroxidase-1 and showed that the efficiency of Sec incorporation can be modulated by UGA position; Sec incorporation at high efficiency appears to require that the UGA be >21 nucleotides from the AUG-start and >204 nucleotides from the selenocysteine insertion sequence element.

Mammalian glutathione peroxidase-1 (GPX1 1 ; H 2 O 2 :GSH oxidoreductase, EC 1.11.1.9) was the first identified selenium-containing enzyme (1), and this discovery provided a biochemical role for selenium. The selenium is present as selenocysteine (Sec or U) incorporated into the peptide backbone, and Sec is located at the active site of GPX1 (2,3). In the intervening years, three additional members of the glutathione peroxidase family have been shown to be selenoenzymes (4 -8). In addition, the three mammalian deiodinases are now known to be selenoenzymes (9); thioredoxin reductase (TR) was recently shown to contain Sec as the penultimate amino acid (10,11), and one isoform of the enzyme responsible for activating selenium, selenophosphate synthetase-2, also contains Sec (12,13). Sec is the active cofactor for these catalytic reactions (3, 14 -16). Finally, selenoprotein W (17) with one Sec and plasma selenoprotein P (18) with up to 10 Sec residues are selenoproteins with unknown function. Selenoproteins and selenoenzymes are also found in the prokaryotic and archae kingdoms (19 -22), showing the ubiquitous distribution of Sec-containing translation products.
For all known selenoproteins, selenium is present as Sec and is incorporated during translation at specific positions encoded by in-frame UGA codons (19,21,(23)(24)(25). A selenocysteine insertion sequence (SECIS) element stem-loop has been shown to be necessary for recognition of the UGA as a Sec codon rather than a nonsense codon (24,26). In eukaryotes, the SECIS element is located in the 3Ј-UTR, whereas in prokaryotes, the SECIS element is located immediately downstream of the UGA codon. Together the UGA and SECIS elements appear to be the two necessary cis-acting mRNA elements that confer recognition of UGA as a Sec codon during translation (24,27,28).
The carbon skeleton of the Sec is provided by serine (29); serine, esterified to a unique tRNA UCA Ser3 Sec in both eukaryotes and prokaryotes, and selenophosphate, synthesized by selenophosphate synthetase, are the substrates used to form Sec-tRNA UCA Ser3 Sec (24,30). This reaction is catalyzed by selenocysteine synthetase (SELA) in bacteria (31,32), but the enzyme is yet to be characterized in eukaryotes. The remaining steps for translational insertion of Sec into selenoproteins must be inferred from the known bacterial reactions wherein a unique elongation factor (SELB) with affinity for the SECIS element promotes the formation of a quaternary mRNA⅐Sec-tRNA UCA Ser3 Sec ⅐SELB⅐GTP complex that leads to translational incorporation of Sec into the nascent polypeptide chain (33,34). Several eukaryotic candidates for SELB have recently been reported (35)(36)(37)(38).
To study Sec incorporation, a good experimental model is needed. 75 Se is readily incorporated into selenoproteins when animals are injected with 75 Se or when cells are cultured with 75 Se (21). Insertion of intact selenoprotein cDNAs or genes into conventional expression vectors, however, typically results at best in 2-5-fold overexpression of GPX1 activity when the host endogenously expresses the selenoprotein (39 -42). Cysteinesubstituted GPX1 has been synthesized in bacteria using mutant Sec3 Cys constructs, but synthesis of eukaryotic seleno-proteins in bacterial systems does not occur due to disparate ribosomal and translational factors (16,43,44). By using transfection in oocytes as well as in several different mammalian cell lines, a number of researchers have studied the effect of various SECIS elements and the effect of primary and secondary SE-CIS structure on Sec incorporation into selenoproteins. SECIS elements from type I deiodinase, GPX1, GPX4, selenoprotein P (with two SECIS elements), and selenoprotein W all confer the ability of an UGA-containing mRNA to incorporate Sec during translation (22,26,27,(45)(46)(47). The efficiency of Sec incorporation occurs at 1/400th to 1/20th of the rate for Cys incorporation, and this efficiency varies markedly depending on the nature of the SECIS element (13,27,45,47). Far fewer studies have directly examined the effect of the UGA position on the efficiency of Sec incorporation. Mutation of four codons to UGA at other positions throughout type I deiodinase demonstrated that a specific UGA context was not required for Sec incorporation (27). Deletion of nucleotide sequences adjacent to the UGA codon of GPX1 resulted in translational incorporation of 75 Se into recombinant GPX1, although relative 75 Se incorporation was reduced for at least four of the constructs (28). One study, which focused directly on the base following the UGA, reported that Sec incorporation could change as much as 3-fold depending on the identity of the fourth base. These reports suggest that additional cis-elements may be required for efficient Sec translation or that the assembly of the required components at the UGA in combination with competition with release factors may be inefficient relative to the incorporation of other amino acids.
Thus we conducted these studies to systematically investigate the effect of UGA position on Sec incorporation. We constructed a genomic GPX1 expression vector that overexpressed 75 Se-labeled recombinant protein at levels greater than the endogenously encoded selenoproteins. Mutation of 10 separate codons to UGA indicated that UGA codons located in the middle of the open reading frame efficiently direct Sec incorporation. Codons located close to the 5Ј-start or 3Ј-termination codons were much less efficient, but insertion of a green fluorescent protein (GFP) coding region as a spacer increased the efficiency of Sec incorporation at these positions to that of the wild-type UGA position. These studies suggest that for optimum Sec incorporation the UGA must be Ͼ21 nt from the AUG-start and Ͼ204 nt from the SECIS element.
GPX1 Expression Vectors-The gpx1 expression vector (Fig. 1a) was constructed by inserting the XbaI/HindIII fragment of the mouse genomic gpx1 gene (2) at the corresponding sites in the multiple cloning region in pRc/CMV vector (Invitrogen, Carlsbad CA) as described previously (42). The f-gpx1 expression construct (Fig. 1b) was prepared by annealing complementary 27-nt oligomers (oligo-57 and -58) encoding the FLAG epitope (48) with SacII restriction site overhangs and ligating this insert into the SacII site of mouse gpx1 in pRc/CMV vector. The construct was confirmed by restriction digestion and by DNA sequencing. The f-gpx1 was then subcloned to the pBluescript II SK(Ϫ) (Stratagene, La Jolla, CA) vector at the XbaI site in the multiple cloning region for use as a template for site-directed mutagenesis.
Mutagenesis-Site-directed mutagenesis was conducted according to standard procedures (49) to change TGA for the native U47 to TGC for Cys. The resulting U47C containing f-gpx1 was used as a common template for changing 10 different codons into TGA by site-directed mutagenesis. These codons were TCC for Ser-7, TCC for Ser-12, TCC for Ser-18, AGC for Ser-29, TCT for Ser-45, TCC for Ser-93, TGC for Cys-154, GTT for Val-172, AGC for Ser-178, and TCC for Ser-195 as indicated in Fig. 1d. Amino acid numbers refer to the original residue position in murine GPX1 (2). All mutants were confirmed by DNA sequencing at the DNA Core Facility at the University of Missouri, Columbia. In total, one wild-type f-gpx1 and 10 mutant f-gpx1 constructs were prepared. Each construct was made and tested independently at least twice.
GFP-GPX1, GPX1-GFP, and GPX1/TAA/GFP Fusion Constructs-The coding region of enhanced GFP was obtained in the pEGFP-N1 vector (CLONTECH, Palo Alto, CA). To make GFP-GPX1 fusion constructs, the F-GPX1 coding region without Met-1 and the GPX1 3Ј-UTR (nt 41-1234) (2) were PCR-amplified with oligo-80 (5Ј primer) and oligo-81 (3Ј primer). Both primers contained BsrgI recognition sequences at the 5Ј end. The amplified fragment was digested with BsrgI and inserted at the BsrgI site of GFP. Clones with GPX1 inserted in sense orientation with GFP were identified by restriction digestion analysis. The resulting GFP-GPX1 constructs had GFP coding region inserted in-frame in front of the F-GPX1 coding region (without Met-1) followed by the GPX1 3Ј-UTR. To make GPX1-GFP fusion constructs, the GFP coding region without Met-1 was PCR-amplified with oligo-104 (5Ј primer) and oligo-105 (3Ј primer). The PCR products were blunted by mung bean nuclease. The F-GPX1 constructs were digested with Bsu36I restriction enzyme, blunted with mung bean nuclease, and ligated with the GFP coding region. Clones with GFP inserted in sense orientation with GPX1 were identified by restriction digestion analysis. The resulting GPX1-GFP constructs had GFP coding region without Met-1 inserted in-frame downstream of the entire GPX1 coding region but before the UAA stop codon. The GPX1/TAA/GFP constructs were made in a similar way as GPX1-GFP except that after the F-GPX1 constructs were digested with Bsu36I enzyme, the TAA overhang was filled in by Klenow fragment. The resulting GPX1/TAA/GFP fusion constructs had GFP coding region without Met-1 inserted immediately downstream from the GPX1 UAA stop codon. For these three sets of fusion constructs, at least two independently constructed fusion genes were examined.
In Vitro Transcription-The 3Ј 477-base pair ApaI/EcoRI fragment of mouse genomic gpx1 was subcloned into pBluescript II SK(Ϫ) vector and then linearized with Bsu36I to make the gpx1 probe template. A 228-base pair EcoRV/PstI fragment of the neomycin resistance gene (neo) was removed from the pRc/CMV vector, subcloned into the pBluescript II SK(Ϫ) vector, and then linearized with SalI to make the neo probe template. In vitro transcription of antisense RNA probes was performed according to the manufacturer's protocol (Promega, Madison. WI) using 40 Ci of [ 32 P]UTP (3000 Ci/mmol) (NEN Life Science Products) and 4 l of 100 M cold UTP per reaction. The full-length probes were purified by electrophoresis through a 6% polyacrylamide gel, eluted in 2 M ammonium acetate ϩ 1% SDS, and ethanol-precipitated.
RNase Protection Assay-Total RNA was isolated from Se-adequate (10 Ϫ7 M Na 2 SeO 3 ) transiently transfected COS-7 cells 48 h after transfection as described previously (41). Two plates of transfected cells were pooled for each sample to provide an adequate yield of RNA. RNA samples from each transfection pool (20 g total RNA) were hybridized overnight at 45°C with the single-stranded antisense RNA probes for mouse GPX1 mRNA and neo mRNA and then treated with RNase (40 g/ml RNase A, 2 g/ml RNase T1) for 45 min at 30°C. After RNase inactivation, the protected probe fragments were ethanol-precipitated, and samples were analyzed on a 5% denaturing polyacrylamide gel. Control samples containing 20 g of yeast tRNA did not protect any detectable probe fragments from the RNase digestion. The experiments were repeated three times with RNA samples from three individual transfections for each construct. Within each sample, the protected GPX1 mRNA signal was normalized to the protected neo mRNA. Analysis of variance was used to determine that GPX1 mRNA levels were not significantly different among COS-7 cells transfected with the various F-GPX1 constructs.
Culture, Transfection, 75 Se Labeling, and Harvesting of COS-7 Cells-COS-7 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 2 mM L-glutamine according to standard cell culture procedures (50). For transient expression, 24 h before transfection, 1 ϫ 10 6 cells were seeded in each 100-mm plate. Twenty micrograms of each plasmid DNA were transfected into cells in each plate by calcium phosphate-mediated method (51). As a control for transfection efficiency, 3 g of pSV-␤-galactosidase control vector (Promega) was co-transfected for each plate. For 75 Se labeling, 5 Ci of 75 Se as Na 2 SeO 3 (100 -400 Ci of 75 Se/g of selenium, pH 7.0, University of Missouri Research Reactor) were added to each plate 24 h after transfection. Forty-eight hours after transfection, cells were harvested and lysed in 900 l of lysis buffer per plate (50 mM Tris, pH 8.0, 150 mM NaCl, 10 mM MgCl, 5 mM EDTA, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 20 g/ml leupeptin). The lysed cell suspension was centrifuged at 14,000 ϫ g at 4°C for 2 min to remove cell debris, and the lysate supernatant was used for subsequent analysis.
Other Assays and Analysis-Protein concentration was quantitated by the method of Lowry et al. (52). ␤-Galactosidase activity was determined with the ␤-galactosidase enzyme assay system according to the manufacturer's instructions (65). GPX1 activity was determined as described (53). Immunoprecipitation was conducted with 50 l of anti-FLAG M2 affinity gel for each aliquot of lysate according to the manufacturer's protocol (Anti-FLAG M2 Affinity Gel, Scientific Imaging Systems, Eastman Kodak Co.). Quantitation of 75 Se incorporation into transfected F-GPX1 was obtained by immunoprecipitation of 200 g of lysate protein followed by counting of the washed precipitate.
Selenoproteins were separated by 10% SDS-PAGE using 0.75-mm gels (51). 75 Se incorporation into selenoprotein bands was quantitated by autoradiography (X-Omat Blue XB-1 film, Kodak) followed by densitometric scanning (LKB 2222 UltraScan, Pharmacia LKB, Uppsala). 75 Se incorporation into transfected F-GPX1 or GPX1 was expressed as a percentage of total 75 Se incorporated into all the selenoproteins. These values were normalized relative to co-expressed ␤-galactosidase activity. For each transfection experiment, the normalized value of each transfected recombinant GPX1 was expressed as a percentage of the level of expression of the wild-type GPX1 in the same transfection experiment. Each construct was examined in at least three independent transfection experiments with at least three replicates for each experiment. Analysis of variance was used to determine differences due to UGA position in mRNA levels and 75 Se incorporation. Differences were considered significant at the probability level p Ͻ 0.05.

RESULTS
Expression of 75 Se-labeled GPX1 in COS-7 Cells-To develop a model selenoprotein to study 75 Se incorporation during translation, the 1497-base pair murine genomic GPX1 sequence was inserted into the pRc/CMV vector (Fig. 1a), transiently transfected into COS-7 cells, and labeled for 24 h with 75 Se. As shown in Fig. 2a, wild-type COS-7 cells prominently label thioredoxin reductase (TR) (10,11), GPX1, and the 18-kDa glutathione peroxidase-4 (GPX4) (5,7,54) as do vector-transfected cells. Cells transfected with GPX1 did not significantly alter 75 Se incorporation into TR or GPX4 when compared with vector-transfected cells, but there was a marked increase in 75 Se incorporation into GPX1. GPX1 activity was increased 2-fold over wild-type cells (data not shown); similar modest increases in selenoenzyme activity are reported for GPX1-transfected cells that possess endogenous GPX1 activity (39 -42). 75 Se incorporation into GPX1, expressed relative to ␤-galactosidase activity to normalize for transfection efficiency, was 3-fold (61 versus 21%) higher than the vector-alone transfectants, showing that this basal construct directed significant 75 Se incorporation into GPX1 in COS-7 cells.
Expression of FLAG-tagged GPX1 (F-GPX1)-To be able to differentiate recombinant GPX1 from endogenous GPX1 in this system, we inserted 27 nt at the SacII site of gpx1 as shown in Fig. 1b to increase the molecular mass of the recombinant protein from 22.3 to 23.5 kDa while adding the FLAG epitope between Ala-8 and Ala-9 of murine GPX1. Expression in COS-7 cells resulted in a new 23.5-kDa 75 Se-labeled protein (U47) (Fig. 2a), indicating that the UGA codon for Sec at original amino acid residue 47 still directed Sec incorporation into the recombinant protein. Expression of this construct had little effect on 75 Se labeling of TR and GPX4, but 75 Se incorporation into the 23.5 kDa F-GPX1 was more than 2-fold higher than 75 Se incorporation into endogenous GPX1 in the same lysate ( Fig. 2a). This model selenoprotein thus could be used to study the effect of UGA position on Sec insertion during translation.
Effect of UGA Position on 75 Se Labeling of F-GPX1-Sitedirected mutagenesis was first used to change the UGA codon for U47 to UGC for Cys (U47C) as shown in Fig. 1c. We next mutated eight Ser, one Cys, and one Val codon to UGA throughout exons 1 and 2 (Fig. 1d). Transient transfection of these constructs all resulted in the expected 75 Se-labeled 23.5-kDa selenoproteins (Fig. 2a). All of the constructs in this series directed 75 Se labeling of a detected recombinant protein. Fig.  2b shows the quantitation of 75 Se labeling by autoradiography and subsequent densitometry, normalized for transfectional efficiency using ␤-galactosidase activity. S93U and C154U, located half-way and three-fourths of the way through the coding region, resulted in significantly higher 75 Se incorporation than for U47. Sec codons within the first 21 nt of the coding region, however, showed dramatically reduced incorporation of 75 Se relative to U47, as did Sec codons in the terminal 90 nt of the coding region (Fig. 2b). Constructs S7U, V172U, and S178U resulted in only 31, 13, and 11% of the normalized 75 Se incorporation for U47 constructs, indicating that UGA codon position can have a dramatic effect on Sec incorporation.
Effect of UGA Position on Immunoprecipitated F-GPX1-The FLAG epitope of these constructs was next used to verify that the 23.5-kDa 75 Se-labeled bands on SDS-PAGE gels were translation products arising from the constructs. This second method assessed independently the effect of UGA codon position on resulting steady-state levels of 75 Se-labeled F-GPX1 in transfected COS-7 cells. As shown in Fig. 3, treatment of 100 g of lysate protein from COS-7 cells transfected with U47 with 6 l of anti-FLAG M2 affinity gel removed the 75 Se-labeled F-GPX1 from the lysate (compare lane 2 with lane 3). SDS-PAGE showed that the protein eluted from the bead was a single 23.5-kDa 75 Se-labeled protein and that the quantity was in proportion to the amount of initial lysate (Fig. 3b). Table I shows that when anti-FLAG M2 antibody was used to evaluate the effect of UGA position on level of 75 Se incorporation, the results were very similar to those determined by SDS-PAGE analysis of 75 Se incorporation into the recombinant GPX1 proteins. For example, immunoprecipitation showed that expression of S7U was 37% that for U47 as compared with 31% for 75 Se incorporation detected by direct SDS-PAGE densitometry. Thus UGA codon position clearly can dramatically affect the expression of this model selenoprotein.
Effect of UGA Position on F-GPX1 mRNA Levels-GPX1 expression in mammalian cells has been shown to be regulated via regulation of mRNA levels (41,42,(55)(56)(57). To determine if the effect of UGA position on detected F-GPX1 proteins in transfected COS-7 cells was due to differences in mRNA level, we used ribonuclease protection analysis to quantitate the levels of the recombinant mRNA in transfected COS-7 cells. The probe, designed for the recombinant murine 3Ј-UTR as shown in Fig. 4, did not protect endogenous COS-7 GPX1 mRNA. The major protected transcript for all constructs corresponds to recombinant GPX1 mRNA that is polyadenylated under control of the genomic GPX1 poly(A) signal rather than the vector bovine growth hormone poly(A) signal. There was no obvious effect of UGA position on steady-state levels of F-GPX1 mRNA.
When the level of protected GPX1 mRNA was normalized to the level of aminoglycoside phosphotransferase (neomycin resistance) mRNA, statistical analysis confirmed that there was no significant effect of UGA position. The trend in the data suggests, if anything, that the level of F-GPX1 mRNA was elevated when the UGA was located at either end of the coding region.
Effect of UGA Position Relative to AUG-Start-The above results suggest the hypothesis that low efficiency of Sec insertion at S7U, S12U, or S18U is because the UGA codons are too close to the AUG start codon. To study this hypothesis, we inserted the GFP coding region, without the AUG, in-frame between Met-1 and Ser-2 of F-GPX1 as shown in Fig. 5. The resulting recombinant 75 Se-labeled proteins were expressed, resulting in the expected 51-kDa size for the GFP-GPX1 fusion proteins. The level of 75 Se incorporation into U47 GFP-GPX1 was almost as high as for F-GPX1 (U47); 75 Se incorporation for UGA codons at S7U, S12U, and S18U was increased relative to the U47 GFP-GPX1 construct. These constructs show that an additional 711 nt between the AUG-start and UGA increases 75 Se incorporation at UGA positions close to the AUG-start.
Effect of UGA Position Relative to UAA-Stop-A parallel GFP insertion construct was used to study the hypothesis that the low efficiency of expression of UGA codons in the terminal 90 nt of the coding region arises because the UGA is too close to the UAA-stop. The GFP coding region, without the AUG, was inserted in-frame immediately before the UAA-stop of F-GPX1 (Fig. 6). The resulting 50-kDa 75 Se-labeled recombinant GPX1- GFP proteins were expressed, but the 75 Se labeling was reduced relative to the endogenous 75 Se-labeled proteins in contrast to the 75 Se labeling for constructs shown in Fig. 2. When the distance between the UGA and the UAA-stop was increased by 711 nt, V172U and S178U GPX1-GFP were expressed at levels virtually the same as for U47 GPX1-GFP. Most interestingly, S195U, which directed Sec incorporation at 43% of U47 in the original F-GPX1 construct (Fig. 2), now directed relative incorporation 4-fold higher than for U47 when the 711-nt insert was added between UGA and the UAA-stop. In addition, a smaller 32-kDa protein was labeled in cells transfected with F-GPX1-GFP S195U; this likely arises from initiation at the Met-142 AUG codon of GPX1. Clearly UGA position can have a significant impact on efficiency of 75 Se incorporation.
Effect of UGA Position Relative to SECIS Element-The above experiment suggests that insertion of a nucleotide spacer between a UGA and the UAA-stop can increase efficiency of 75 Se incorporation, but this insert also increases the number of nucleotides between the UGA codon and the SECIS element. Thus we made a second series of constructs with the 711-nt GFP insert positioned immediately after the UAA-stop. Fig. 7 shows that transfection of COS-7 cells with these GPX1/TAA/ GFP constructs resulted in recombinant F-GPX1 of the same size as for the F-GPX1 constructs shown in Fig. 2. This permitted direct comparison of relative translational efficiency. As shown in Fig. 7, a 711-nt insertion between UAA-stop and the SECIS increased 75 Se incorporation into V172U and S178U constructs, indicating that the spacing requirement was for additional bases between the UGA and the SECIS and not between the UGA and the UAA-stop. As with the GPX1-GFP S195U construct, the 711-nt insert increased 75 Se incorporation in GPX1/TAA/GFP S195U over 3-fold relative to total 75 Se incorporation by these transiently transfected cells; this is almost 10-fold relative to the level of F-GPX1 S195U. Interestingly, this GFP insert after the UAA adversely affected 75 Se incorporation for the U47 and S93U constructs, suggesting that the optimum incorporation of Sec at UGA codons in the middle of native GPX1 may also reflect an optimal spacing between the UGA codon and its tethered SECIS element. DISCUSSION These studies used model recombinant gpx1 genes to study the effect of UGA position on translation as measured by 75 Se incorporation into the encoded protein. Transiently transfected COS-7 cells cultured in 75 Se-containing media incorporated 75 Se into recombinant GPX1 such that 75 Se in GPX1, both transfected and endogenous, was 61% of the total 75 Se-labeled proteins as compared with 21% for cells transfected with the vector alone. When a 27-nt insert encoding the FLAG epitope was inserted eight codons from the AUG-start, the resulting larger F-GPX1 was expressed and consistently accounted for 30 -36% of the total 75 Se incorporated into selenoproteins. This high level of 75 Se incorporation into the recombinant GPX1 or F-GPX1 shows that mRNA transcribed from this model recombinant gene is efficiently translated, at rates comparable to endogenous selenoproteins. Thus this system was used to study the effect of UGA position on 75 Se incorporation.
When the position of the UGA in the F-GPX1 construct was varied across the full coding region, the level of 75 Se-labeled F-GPX1 increased for S93U and C154U relative to wild-type U47 and S45U, showing that Sec-encoding UGA codons in the middle of the open reading frame readily direct Sec incorporation and that there is nothing unique about the U47 position. Berry et al. (27) found that single UGA codons near the middle of the coding region were used for Sec translation with similar efficiencies to UGA at the wild-type position in deiodinase. When we used site-directed mutagenesis to convert three Ser codons into UGA codons, located 21, 63, or 81 nt from the AUG-start, 75 Se incorporation into F-GPX1 was reduced relative to U47. Addition of a 711-nt GFP insert in-frame between the AUG-start and the first UGA, however, increased 75 Se incorporation at all three positions. The GFP insert increased 75 Se incorporation at S7U from 31 to 79% relative to U47. This experiment indicates that UGA codons minimally positioned within the first 21 nt do not efficiently direct 75 Se incorporation into selenoproteins.
When three codons, positioned 21, 72, or 90 nt from the UAA-stop codon, were mutated to UGA codons, 75 Se incorporation into the resulting recombinant protein was also reduced to 13, 11, or 43%, respectively, of that determined for U47. Insertion of the 711-nt GFP coding region immediately before the UAA-stop increased the directed 75 Se incorporation to levels equivalent to those for the corresponding allele with UGA at position U47. Insertion of the 711-nt GFP immediately after the UAA-stop, however, also increased 75 Se incorporation. These experiments show that the distance from the SECIS element, rather than the UAA-stop, is important for UGA to direct efficient translational insertion of 75 Se. These results indicate that UGA codons placed closer than 150 nt are not as well translated as codons 204 nt from the SECIS element.
We have used the level of 75 Se-labeled recombinant GPX1 protein to determine the effect of UGA position on 75 Se incorporational efficiency during translation. Importantly, ribonuclease protection assays (Fig. 4) showed that UGA position had no effect on level of recombinant mRNA. Within each series, we selected small neutral amino acids for Sec substitution, thus minimizing potential changes on tertiary structure and thus a 75 Se-Labeled cell lysates from COS-7 cells transiently transfected with f-gpx1 with Sec (U) at indicated positions were analyzed by SDS-PAGE. After densitometric scanning of the autoradiogram, 75 Se incorporation into each recombinant F-GPX1, calculated as a percentage of 75 Se incorporation into total selenoproteins, was normalized relative to co-expressed ␤-galactosidase activity. Normalized values for different mutants were expressed as a percentage of the value of F-GPX1 U47 transfected in the same experiment.
b Values for each construct are the mean Ϯ S.E.M of at least three independent transfection experiments. c 75 Se-Labeled cell lysates from COS-7 cells transiently transfected with f-gpx1 with Sec (U) at indicated positions were analyzed by immunoprecipitation. Lysates (200 g of protein) from cells transfected with f-gpx1 with Sec at indicated positions were treated with 50 l of anti-FLAG M2 antibodies, and immunoprecipitates were washed and then counted for 75 Se. Precipitated 75 Se counts were normalized relative to co-expressed ␤-galactosidase activity and expressed relative to immunoprecipitated 75 Se from F-GPX1 U47 transfected cells.
protein turnover. Thus, within a series of constructs, levels of 75 Se-labeled protein reflect the relative translational efficiency. This approach has been used by other researchers. Shen et al. (28) used 75 Se incorporation into recombinant GPX1 to evaluate efficiency of 3Ј-UTR sequences for directing Sec incorporation. Martin et al. (46) used 75 Se-labeled recombinant proteins, detected by SDS-PAGE, to compare translational efficiency of recombinant type I deiodinase constructs as modulated by SE-CIS location, and they found that 75 Se-labeled protein levels matched with deiodinase activities. Leonard et al. (58) used 75 Se labeling of recombinant selenoproteins, detected by immunoprecipitation, and found that 75 Se labeling correlated with antibody binding. Thus the level of 75 Se-labeled protein can be used routinely to quantitate translation of selenoproteins. Similarly, deiodinase activity or substrate binding, requiring Sec incorporation at UGA to synthesize the full-length protein, was used quantitatively to identify the SECIS element (26), to compare efficiency of various SECIS elements for directing Sec insertion (27,59), and to determine the essential components of SECIS secondary structure (27,46). Collectively our experiments and previous studies show that the level of recombinant protein, whether assessed by 75 Se labeling, enzyme activity, immunological reactivity, or substrate binding, can be used to monitor efficiency of translation.
Protein synthesis requiring Sec translation is reported to be far less efficient than for Cys mutants (19). Incorporation of Sec into deiodinase-1 directed by a UGA codon is reported to be 5 to 0.25% of the rate for Cys incorporation (45). More recently, Kim et al. (13) reported that Sec incorporation into selenophosphate synthetase-2 occurs at 2.5% the rate for Cys incorporation. Experiments with a novel ␤-galactosidase-luciferase expression vector with UGA or UGC in-frame between the two reporter genes found that Sec incorporation was 2.8% the rate for Cys incorporation (47). Thus Sec translational efficiency directed by native UGA codon position is reduced relative to standard amino acid translation. And more than a UGA and a SECIS element appear to be required for Sec translation, as addition of a UGA codon and a GPX1 3Ј-UTR to the ␤-galactosidase coding region in an expression vector did not result in expressed 75 Se-labeled protein in our multiple initial attempts to develop a ␤-galactosidase model selenoprotein expression system for these experiments (data not shown).
Our results indicate that the translational efficiency of UGAmediated Sec incorporation diminishes when the UGA-SECIS distance is 150 nt or less. Berry and colleagues (46) found that there is a minimal spacing distance of ϳ60 nt necessary so that the tethered SECIS is capable of catalyzing Sec insertion. They found that a spacing of 51 nt between the UGA codon and the SECIS element (the boundary of the conserved 5Ј AUG) resulted in almost complete loss of expression of recombinant deiodinase (46). Most selenoproteins do not use UGA as the stop codon, but selenoprotein W can use a second UGA as the stop presumably because the 55-57 nt between the UGA-stop and the SECIS effectively terminates translation (60). These results indicate that there is a minimum permissive UGA-SECIS distance for Sec incorporation versus termination; our results indicate, in addition, that a 204-nt distance is required for optimum translational efficiency. This distance might be one of the ways that biology regulates relative ability of various mRNA species to promote Sec incorporation.
An alternative explanation for our results is that the base immediately following the UGA may play a role in efficiency of Sec translation. McCaughan et al. (61) suggest that when the 4th base is U or C, Sec incorporation is enhanced 3-fold in comparison to the 4th bases of G or A. Our UGAs at positions S29U, S45U, S93U, C154U, V172U, S178U, and S195U all are followed by C and yet have widely differing Sec translation efficiencies. Similarly, wild-type U47, S7U, and S18U all are followed by G. Thus translational termination efficiency as influenced by the 4th base does not provide a ready explanation for the effect of UGA codon position on Sec translational efficiency.
Construct S195U in these experiments suggests that additional features are important in determining the full ability of a UGA to direct Sec incorporation. In the F-GPX1 construct series, S195U directed 75 Se incorporation at 43% of the rate of U47. With a spacing of 81 nt between the UGA and the SECIS, this reduced efficiency is consistent with our proposal that Sec translational efficiency is decreased under these conditions. This rate, however, is 4-fold higher than for S178U with a spacing of 132 nt. This suggests that S195U may have a unique context that enhances Sec incorporation despite its adjacency to the SECIS element. This is affirmed by the GFP insert constructs that resulted in 10-fold increases in Sec insertion at S195U. In addition, the novel 32-kDa 75 Se-labeled second protein product was observed with several independent constructs, in contrast to the lack of any other similar 75 Se-labeled products with any of the other constructs. The alternative start at the Met-142 AUG should permit 75 Se incorporation at downstream UGAs located at V172U or S178U, but the resulting 32-kDa 75 Se-labeled proteins for these constructs were not detected (Fig. 6), further illustrating the importance of UGA position and context in affecting Sec incorporation.
The explanation for the decreased efficiency of Sec incorporation at UGA codons close to the 5Ј-start or 3Ј-stop is not clear. As discussed above, a spacing of 204 nt between UGA and the SECIS for optimum efficiency may just be an expansion of the minimum spacing requirement of Berry and co-workers (46) that provides optimum flexibility so the SECIS element can interact effectively with the elongation complex. An explanation for the reduced efficiency of UGA codons adjacent to the 5Ј-start, however, is less clear, as 1.5-kilobase inserts or SECIS elements in trans are reported to still facilitate Sec incorporation (27). The studies presented here were conducted with the murine GPX1 SECIS element; SECIS elements from different selenoproteins confer different efficiencies for Sec incorporation (27), so different SECIS elements might also have disparate effects on efficiency of Sec insertion at codons close to the AUG-start. One hypothesis for these results is that the assembled 48 S initiation complex (62) competes with proper assembly of the ternary Sec incorporation complex when the UGA codon is within 21 nt of the AUG-start. A requirement of 16-nt spacing between the 5Ј-AUG and the UGA has been demonstrated for prokaryotic Sec incorporation (63). The recent suggestion of a poly(A)-binding protein-mediated physical link between the termini of the mRNA during translation (62,64) could extend these potential interactions to explain decreased efficiency when UGA codons are located at either extreme of the open reading frame. An alternative hypothesis is that there is a third element or feature of selenoprotein mRNAs, perhaps involving UGA context or RNA secondary structure, that has been optimized for native UGAs in selenoprotein mRNAs. This would explain why the presence of a UGA and a SECIS alone is not sufficient for efficient expression of every recombinant mRNA that contains these two cis-elements.
These experiments show that the translational efficiency of UGA codons to direct Sec incorporation during translation is modulated when the UGA codons are located within a minimal distance of the AUG-start or within a minimal distance of the SECIS element. Sec incorporation at high efficiency appears to require that the UGA be Ͼ21 nt from the AUG-start and Ͼ204 nt from the SECIS element. Dramatically enhanced Sec translational efficiency of S195U when the distance constraint is eliminated suggests that there can be additional components, such as UGA context or local RNA secondary structure, that can modulate the efficiency Sec incorporation during translation.